I'm vaguely remembering a talk I saw at Microsoft Research in 2013 by someone who studies psychoacoustics at a university I can't recall. They created an acoust

I'm vaguely remembering a talk I saw at Microsoft Research in 2013 by someone who studies psychoacoustics at a university I can't recall. They created an acoustic model that predicted certain qualities of a sound (wetness, harshness, brilliance, etc.), trained with data where subjects assigned such qualities on a scale for many example sounds. They used this to create 2D synth controls using the salient parameters. I believe this ended up being a somewhat popular VST plugin.